I have a little theory that can't be proven or disproven (a great advantage) but which at least reminds us to be cautious.
After a headline-grabbing show-case political event such as the annual conference of a party, you usually get a small sympathetic move in the polls, which tends to disappear a few days later - it happens after the Democratic and Republican conventions in the US as well. This doesn't surprise us, but perhaps it should: do we really imagine that significant numbers of people change their minds about who to vote for after a single staged event? And then reverts?
It seems unlikely. More likely, I believe, is that those who take least interest in political matters, but who feel they should give the pollsters an opinion other than 'don't know', are more likely to be swayed by whatever they happened to hear most recently. And are least likely actually to vote.
Think of that after the debates. It is of course possible that one candidate or other will make such a strong or hopeless performance that floating voters flip from one side to another. But for that to happen, we would have to suppose a very large, highly engaged and yet strongly personality-influenced audience.
It strikes me as more probable that the majority of the audience have already committed and are watching as a spectator sport. Short-lived mini-bounces in the polls are therefore much more likely to come from people who don't really care but are picking up a little buzz from the papers the next day.
That, of course, is still worth noting; under the froth there may yet be a little substance; we just shouldn't over-interpret its significance.
